Dithering is a way to get rid of sensor irregularities (hot pixels, pattern noise, etc) by moving your target to different parts of the sensor. This way, the target accumulates while stacking, but the sensor irregularities moves around, and averages out to become smaller errors. Roughly speaking, (error + no error + no error + no error) would result in 0.25x error power, for example, while (target + target + target + target) would result in 4x target power. There is still the same amount of total noise, but they are now spread all over the image, so they are not as visible anymore.
So, what you want to do is to dither enough times per total number of frames. If you are going to accumulate 200 subframes, you probably won't need to dither more often than once per 10 frames. Since a bad pixel would have moved to 20 different locations relative to your target.
However, if you are only gathering 10 sub-frames, you would probably want to dither after every frame, otherwise the dither won't be very effective in getting rid of the gross imprefections of the sensor.
So, how often you dither is dependent to how many frames you are going to end up stacking, and not dependent on the exposure duration of each subframes.
If you are using a Bayer ("color") camera, I would recommend a movement that is more than 2 pixels each time to further minimize the effect of a hot or a dead pixel. If you think the dithering mechanism is accurate (the ASIAIR is not nessarily accurate -- I don't know if they have ever fixed it), dither a color camera with odd numbers of pixels. That way you don't see a trail of red pixels, for example, but a mish mash of red, green and blue bad pixels, and those are three times easier to get rid of.
The only drawback of large dither movements is that it takes longer to recover from a dither, and you need to crop a little more of the edges. If you are not concerned about settling time, and have a large enough sensor, by all means use larger dither movements.
Just experiment with it, and you will learn from your experiments. Try not dithering first to get a baseline. What you probably find with a color camera is that the warm pixels will show up as bright red, green and blue dots (tons of them in a camera like the ASI533MC, ASI2600MC and ASI6200MC). Now try dithering every 10 images... you will find that the errors are less bright and visible. Now try dithering every image, and you will find the errors to be even less visible. You can then pick how much warm pixel noise you can tolerate vs the time lost for dithering. The same for other pixel imperfections. Just slowly learn. Astronomy is not a hobby for the impatient. I am now on my 60th year of doing it, and still learning.
Chen
